10 DSC: Lymphoma vs GBM
I would like to differentiate between “Glioblastoma” vs “Primary CNS lymphoma” using advanced brain MRI technique such as DSC (rCBV), DCE (k-trans), and MR spectroscopy.
In one line: Combine perfusion and metabolic fingerprints—low rCBV + very high Ktrans/extraction-fraction + prominent lipid/lactate peaks in a non-necrotic mass → PCNSL, whereas high heterogenous rCBV ± ring necrosis + only moderate permeability + variable lipids confined to necrotic core → GBM.
10.1 1 | Dynamic Susceptibility-Contrast (DSC) Perfusion
10.1.1 How to acquire
- Pre-load + leakage-corrected processing to neutralise T1-shine-through, then normalise the max-rCBV to contralateral white matter. (PMC)
- Place a small ROI on the highest-CBV “hot spot” (avoid visible vessels/necrosis) and a larger ROI for mean-rCBV; record the percentage-signal-recovery (PSR) curve as well. (EJ Radiology)
10.1.2 Typical findings
| Metric | Glioblastoma (GBM) | Primary CNS Lymphoma (PCNSL) | Practical cut-off |
|---|---|---|---|
| Max- or mean-rCBV | 3–6 (mean ≈ 4.9 ± 2.2) (PubMed) | 0.8–2.0 (mean ≈ 1.7 ± 0.6) (PubMed, PMC) | < 1.5–1.7 ⇒ strongly favours PCNSL (PMC) |
| PSR (signal recovery) | Low/variable (due to neo-vascular leak) (PMC) | High PSR (> 90 %)—reflects slow wash-out and intact lumina (PubMed) | |
| Extraction fraction (from DSC leakage models) | Lower | Higher—even in hyper-vascular outliers (ajnr.org) | > 0.25 suggests PCNSL |
Why? GBM shows florid angiogenesis with many patent vessels, hence high blood volume; PCNSL grows angiocentrically, destroys endothelium, but recruits few new vessels → low CBV yet marked contrast leak. (PMC)
10.2 2 | Dynamic Contrast-Enhanced (DCE) Perfusion
10.2.1 Acquisition pearls
- 3-D spoiled gradient-echo every 2–4 s for ≥5 min; Tofts model for Ktrans, ve, Kep.
- Use the same ROI strategy as DSC and map peritumoral rim separately. (PubMed)
10.2.2 Pharmacokinetic signatures
| Parameter | GBM | PCNSL |
|---|---|---|
| Ktrans (min⁻¹) | 0.2–0.6 (moderate) (PubMed) | 0.3–0.8 (often higher) (PubMed, PubMed) |
| Kep (min⁻¹) | High in peritumoural zone (disrupted BBB) (PubMed) | Lower |
| ve | Larger heterogeneous extracellular space | Smaller, homogeneous |
| Diagnostic ROC | Kep peritumoural AUC ≈ 0.90; Ktrans core AUC ≈ 0.82 (PubMed, ajnr.org) |
Interpretation tip: a mass with low rCBV but high Ktrans is classic for lymphoma; high rCBV with only moderate Ktrans suggests GBM.
10.3 3 | Proton MR Spectroscopy (¹H-MRS)
| Spectral marker | GBM | PCNSL |
|---|---|---|
| Cho/Cr ratio | 2–4; elevated in tumour & necrotic rim | > 2.6 (mean ≈ 3–4) (PMC) |
| Lipids/Lactate | Prominent only in necrotic core | Massive peaks even in solid, non-necrotic tissue (PubMed, The Journal of Neurosurgery) |
| Glu/Cr | Lower | Higher (cut-off ≈ 2.5) (PubMed) |
| NAA | Markedly decreased | Markedly decreased |
| Myo-inositol | Variable | Often low |
Pearls
- A large lipid peak without frank necrosis is almost pathognomonic for PCNSL. (The Journal of Neurosurgery)
- Elevated glycine & alanine favour GBM but are not specific.
10.4 4 | Step-by-Step Workflow for Unknown Enhancing Mass
Conventional MRI: if homogeneous enhancement + diffusion restriction → think PCNSL; if ring-necrosis → think GBM.
DSC
- rCBV < 1.5 → go to DCE/MRS to confirm lymphoma.
- rCBV ≫ 2 → GBM more likely; still do DCE for atypical cases.
DCE
- High Ktrans / extraction fraction with low vp → PCNSL.
- Moderate Ktrans with high vp → GBM.
MRS
- Lipid/Lac peak outside necrosis + Cho/Cr > 2.6 → PCNSL.
- Only necrotic-core lipids, Cho/Cr ~ 3, ± glycine → GBM.
Integrate DWI/ADC (low ADC in PCNSL) and clinical factors (age > 60, immunosuppression).
10.5 5 | Pitfalls & Practical Tips
- Steroids rapidly lower lymphoma rCBV & enhancement—perform imaging before high-dose steroid therapy. (PMC)
- Hemorrhage / calcification cause susceptibility drop-outs on DSC—use spin-echo or multi-echo alternatives. (ScienceDirect)
- Hyper-vascular PCNSL (~10 %) can mimic GBM; permeability metrics (extraction fraction, Ktrans) rescue specificity. (ajnr.org)
- Always normalise to contralateral white matter to reduce scanner-to-scanner variability. (PMC)
10.6 6 | One-Glance Summary Table
| Technique | Key Metric | GBM (Typical) | PCNSL (Typical) | Cut-off / Clue | Most Common Pitfall |
|---|---|---|---|---|---|
| DSC | Max/mean rCBV | 3 – 6 (heterogeneous) | 0.8 – 2 (uniformly low) | rCBV < 1.5 ⇒ PCNSL | Hyper-vascular PCNSL subset |
| PSR | Low | High | High PSR (>90 %) ⇒ PCNSL | Uncorrected leakage | |
| DCE | Ktrans (core) | 0.2 – 0.6 | 0.3 – 0.8 | High with low rCBV ⇒ PCNSL | Low temporal resolution |
| Kep (rim) | High | Lower | AUC 0.90 (Kep) | ROI mis-placement | |
| MRS | Lipid/Lac peaks | Necrotic only | Diffuse, intense | Non-necrotic lipids ⇒ PCNSL | Short TE required |
| Cho/Cr | 2 – 4 | > 2.6 | Threshold 2.6 | Partial-volume CSF |
10.6.1 Take-home
When rCBV and metabolic data disagree, believe permeability & lipids—a “leaky, lipid-rich, but low-vascular” tumour is lymphoma until proven otherwise. Conversely, a “vascular, necrotic, cho-rich” lesion points toward glioblastoma. Apply the whole toolkit systematically to maximise diagnostic confidence and to guide biopsy planning or upfront chemoradiation.